Hacker News Reader

Doesn't this harm the claim that the models do not contain any information about a specific image by an artist?

I think there's probably a big difference between how data is stored in art models versus how LLMs store textual information.

If so, maybe that says something about that claim?